HackerTube 中文?
Representation, intuition and feedback →

After an edit, how can you see what actually changed?

A green test, a live picture and a replayable history all count as feedback. Follow one edit and ask which consequences become visible and which still need reflection.

Questions in this reading · 12
  1. How quickly can a programmer see a mistake?
  2. What if feedback arrived with the edit?
  3. Can creators see the effect of an edit?
  4. What can you stop imagining once an edit is visible?
  5. How can an environment make time visible?
  6. Must we choose between reflection and instant feedback?
  7. From a single edit to a usable prototype
  8. Does a local replay actually start from the same frame?
  9. Did the avatar send the right expression in time?
  10. Why handle a component if you can order it the same day?
  11. After seeing a failure, how soon can you try again?
  12. Why is the page still stuck after using asynchronous callbacks?

How quickly can a programmer see a mistake?

Running tests on every file change shortens one feedback loop. Showing the changing behavior beside the edit shortens another. The signals tell the programmer different things.

Compare original passages and timestamps · 2
each time something changes in the directory it reruns the command
A Tour of the Acme Editor · ▶ 20:20–20:39
if you make a change or you make a decision you need to see the effect of that immediately
Bret Victor - Inventing on Principle · ▶ 2:06–2:18

What if feedback arrived with the edit?

One talk traces a search for shorter feedback loops. The other shows what creators can discover when the result is continuously visible.

Compare original passages and timestamps · 2
What if we shortened the feedback loop to effectively nothing
In Search of Tomorrow · ▶ 4:08–4:11
creators need an immediate connection to what they create
Bret Victor - Inventing on Principle · ▶ 2:01–2:08

Can creators see the effect of an edit?

A notation tool and a programming environment both make an edit visible. Look for the precise gap each closes between an action and the result its author can inspect.

Compare original passages and timestamps · 2
...by making the arrow a subtle shade of blue? Do you know how hard that is to see?
Music Software & Bad Interface Design: Avid’s Sibelius · ▶ 19:04–19:10
you need to see the effect of that immediately
Bret Victor - Inventing on Principle · ▶ 2:08–2:16

What can you stop imagining once an edit is visible?

Two programming environments and a notation editor let their users inspect the effect of an edit. Feedback reduces mental simulation without guaranteeing conceptual understanding.

Compare original passages and timestamps · 3
if you make a change or you make a decision you need to see the effect of that immediately
Bret Victor - Inventing on Principle · ▶ 2:08–2:16
So I could literally watch the data flow through my program.
In Search of Tomorrow · ▶ 4:56–5:01
...by making the arrow a subtle shade of blue? Do you know how hard that is to see?
Music Software & Bad Interface Design: Avid’s Sibelius · ▶ 19:04–19:10

How can an environment make time visible?

One demonstration makes behavior visible while editing on a screen. Another brings that feedback into a physical workspace. Compare what the maker can inspect in each setting.

Compare original passages and timestamps · 2
immediately you have to map time to space
Bret Victor - Inventing on Principle · ▶ 13:38–13:47
look across a range of time at once
Bret Victor - Seeing Spaces · ▶ 8:15–8:21

Must we choose between reflection and instant feedback?

Thinking away from the keyboard helps frame the problem. A responsive medium helps inspect an idea once expressed. The two techniques address different moments in the work.

Compare original passages and timestamps · 2
you're going to try stuff and go back
Hammock Driven Development - Rich Hickey · ▶ 37:44–37:51
creators need an immediate connection to what they create
Bret Victor - Inventing on Principle · ▶ 2:01–2:08

From a single edit to a usable prototype

The Cursor team wants a suggestion to arrive before the next edit. Levels works at a different granularity: as a sole operator he can shorten the interval between writing and shipping. Kay’s Alto made a future machine’s interactions usable early, on an expensive stand-in. These loops test different predictions: the next edit, demand for a shipped feature, or the viability of an interaction model. Fast completion cannot answer the third question.

Passages and timestamps · 3
predicting what the user’s going to do next in their editing of existing code
Cursor Team: Future of Programming with AI | Lex Fridman Podcast #447 · ▶ 18:42–19:22 · passage interval
being alone by myself on my laptop, in my underwear in a hotel room or something, I can ship very fast
Pieter Levels: Programming, Viral AI Startups, and Digital Nomad Life | Lex Fridman Podcast #440 · ▶ 3:22–4:02 · passage interval
it was there to simulate this laptop of the future
Founder School Session: The Future Doesn't Have to Be Incremental · ▶ 38:22–38:53 · passage interval

Does a local replay actually start from the same frame?

Sanderson repeatedly runs a small part of a Manim animation to tune it. Pasting the same code again would run it on a scene already changed by the first pass. His replay restores the state at the section’s start. A cheap check for a live preview is to replay twice without editing. If the images already differ, the preview cannot yet isolate the effect of the next edit.

Passages and timestamps · 1
It has to revert to the state that the scene had at the start.
How I animate 3Blue1Brown | A Manim demo with Ben Sparks · ▶ 6:12–6:15

Did the avatar send the right expression in time?

Zuckerberg explains how scans across expressions build an avatar that can then be driven by an encoded signal during a call. Carmack supplies another acceptance condition: subtle reactions must arrive in time for two people to feel together. Put a raised eyebrow, its display and the partner’s response on one timeline. A still frame checks the reconstructed expression. A back-and-forth exchange tests delay. A more convincing frame leaves that second question open.

Passages and timestamps · 4
scanned ourselves in a lot of different expressions
Mark Zuckerberg: First Interview in the Metaverse | Lex Fridman Podcast #398 · ▶ 3:03–3:09
send an encoded version of what you're supposed to look like over the wire
Mark Zuckerberg: First Interview in the Metaverse | Lex Fridman Podcast #398 · ▶ 3:28–3:34
the super low latency responsiveness
John Carmack: Doom, Quake, VR, AGI, Programming, Video Games, and Rockets | Lex Fridman Podcast #309 · ▶ 209:56–209:59
just occupying the same field of view
John Carmack: Doom, Quake, VR, AGI, Programming, Video Games, and Rockets | Lex Fridman Podcast #309 · ▶ 210:01–210:03

Why handle a component if you can order it the same day?

Kay’s flight example asks how soon a failed experiment can be tried again. Shenzhen’s component markets add the waiting before a trial begins. One entrepreneur describes some locally ordered parts arriving that afternoon or the next day; another explains why photographs and specification sheets are insufficient for choosing a switch: feel, size and mounting details need direct inspection. Nearby parts reduce waiting and expose a poor choice earlier; a suitable model can then be purchased in quantity in the same market. Separate these steps to distinguish faster procurement from earlier testing of a product assumption. The documentary’s later discussion of rents and urban villages adds a dependency: these conditions require suppliers, workspace and local experience together. A faster tool for an individual team cannot simply reproduce that network.

Passages and timestamps · 2
we can order in the morning and then we get it in the afternoon
Shenzhen: The Silicon Valley of Hardware (Full Documentary) | Future Cities | WIRED · ▶ 9:50–9:57
you want to know how it feels
Shenzhen: The Silicon Valley of Hardware (Full Documentary) | Future Cities | WIRED · ▶ 26:25–26:29

After seeing a failure, how soon can you try again?

An immediate display shortens only part of a feedback loop. Kay’s account of human-powered flight contains a slower step: after a prototype crashes, will repair consume months? In his telling, MacCready first sought a design that could fail repeatedly in a day and be repaired in minutes, making more informative trials possible. Bring that question back to programming tools. Alongside time-to-result, measure time-to-next-experiment. Reversible edits and recoverable test environments may remove a bottleneck before faster screen updates do. This is a design question drawn from the case, not a law Kay established for every research process.

Passages and timestamps · 2
we need to be able to do 12 crashes a day
Alan Kay - Normal Considered Harmful · ▶ 55:37–55:40
you could fix it in a few minutes
Alan Kay - Normal Considered Harmful · ▶ 55:52–55:58

Why is the page still stuck after using asynchronous callbacks?

Draw “work has been queued” and “the user has seen the result” as two separate events. In Roberts’s sequence of timers, later callbacks still wait behind earlier work, so a one-second delay is an earliest-execution condition, not a promise to execute at one second. His scrolling example adds another failure mode: events enqueue quickly while slow work accumulates in the queue. Use that model to trace a feedback tool through input, queueing, execution and rendering. Replacing a synchronous loop with callbacks does not automatically shorten the whole path. The simplified rendering queue in his animation exposes waiting points; it is not a complete specification of every browser’s scheduler.

Passages and timestamps · 2
it's a minimum time to execution
What the heck is the event loop anyway? | Philip Roberts | JSConf EU · ▶ 20:02–20:07
you're flooding the queue with queued events.
What the heck is the event loop anyway? | Philip Roberts | JSConf EU · ▶ 25:22–25:29

Try it on your own example

Take a recent edit. Which intermediate state would disappear if you saw only the final frame?